id
uo € oM
RP CONES OG I--———
NE CC
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
One more step was conducted to determine the starting values
used for future growth simulation. The best values identified
above were actually the starting values. Because of self
modification incorporated in the model, these starting values tend
to be altered when a run is completed. Thus, a coefficient may
have different starting and finishing values for each run. At the
end, the final values ofthe control coefficients are: diffusion(71 »
breed(10), spread(32), slope resistance(73), and road
gravity(100).
It should be pointed out that the model calibration was carried out
with the use of 240 m resolution dataset only. An earlier test
estimates that the time for completing the first stage of calibration
using the 120 m resolution data set would be about 32,500 hours
or 135 days given the computer resource available. For practical
reason, the other two higher resolution data sets were not used in
model calibration.
2.3 Scenario Design and Simulation
Two possible planning scenarios for future urban development in
Atlanta were considered here, which are tied with different
policies and environmental conditions.
2.3.1 Scenario One: This scenario assumes the factors for the
growth remain unchanged and thus, it may be termed as
‘continuation’. It provides therefore a benchmark for comparison
with the other alternative growth strategy. To implement this
scenario in model simulation, the values of growth control
coefficients obtained from the model calibration were used as the
starting values. The 1999 urban extent data was actually used in
the simulation and other conditions and input data set can be
found from Table 2.
Table 2 The conditions applied for each simulation
Calibration Runs Parameter
Hem Coarse | Fine | Final |Averaging
s resolution (m) 240
à urban extent (year) 1973. 1979. 1987. 1993. 1999
2 “roads” (year) 1973. 1987. 1999
= excluded area stream buffered zones not considered
slope same (only one layer is available)
hillshaded relief same (only one laver is available)
E seed* 2840, 2840 2840 2840
E. number of times** 4 6 10 100
t diffusion coeff start 0 40 52 55
SS | diffusion coeff step 25 5 3 I
2 diffusion coeff stop 100 60 58 SS
breed coef! start 0 0 2 8
breed coeff step 25 5 3 I
breed coeff stop 100 10 8 8
spread coeff start 0 5 22 25
spread coelT step 25 3 3 I
spread coelf stop 100 35 28 25
slope resistance start 0 40 47 53
slope resistance step 25 5 3 1
slope resistance stop 100 SS 53 53
road gravity start 0 80 90 100
road gravity step 25 10 5 |
road gravity stop 100 100 100 100
Number of iteration(s) 3125 900 243 |
* The definitions are: critical high : when the growth rate exceeds this
value, self-modification increases the control parameter values;
critical low: when the growth rate falls below this value,
self-modification increases the control parameter values); boom: value of
the multiplier (greater than one) by which parameter values are increased
when the growth rate exceeds critical high; bust: value of the multiplier
(less than one) by which parameter values are decreased when the growth
rate falls below critical low), and Critical slope: average slope at which
system increases spread.
** Both starting and ending values are given. It should be noted that the
ending values were the averaged values after100 times of Monte Carlo
computations.
*** Program code was changed to allow up to 200 for road gravity.
2.3.2. Scenario Two: The second scenario considers a hybrid
growth strategy in which both conventional suburban development
and alternative growth efforts, such as smart growth and new
urbanism, are addressed. This scenario also considers
environmental conservation by limiting development
around several predefined buffer zones.
To implement this idea in model simulation, the starting values for
five growth control coefficients used in the first scenario need to
be changed in order to slow down the growth rate and to alter the
growth pattern. The conditions used in this scenario can be seen
from Table 2. Please note that the proposed transportation
improvements and new additions as well as environmental
conservation introduced in the second scenario are still valid here.
Although the two scenarios are different in policies and
environmental conditions, there are several commonalities. The
time span is the same, which is from 2000 to 2050. Because ofthe
limitation in computation resources, only the data set with 240 m
spatial resolution is used. The two input data layers, namely,
slope and hillshaded relief, are used without change for all the
runs. The number of times of Monte Carlo computations is 100
and the random samples are 2,840, or about 1 percent of the total
pixels available.
3. RESULT
The progressive urban development as projected into the future 51
years under two different scenarios can be perceived quite well
from Figure 2. The graphical outputs ofthe two scenarios are quite
similar. By evaluating these graphical outputs carefully, it is found
that a Los Angeles-like metropolis characterized by huge urban
agglomerations would emerge by 2030, if current development
conditions are still valid. The vegetation area and open space in
the 13 metro counties (excluding the northwestern mountainous
area) would be very limited. In contrast, the simulated
urbanization under the second scenario appears to be relatively
restrictive, indicating that the effort of slowing down urbanization
through model parameterization has been quite efficient.
Statistical measures reveal much more information. Under the first
scenario, the total urban area for 2050 would be 1,286,692 ha.
The total net increment in urban area with at least 5096 probability
would be 793,561 ha., or 43.6 ha. per day on the average,
representing an increase of 160 percent between 1999 and 2050.
As a result of such a dramatic growth, urban land would occupy
approximately 78.67 percent of the total modeled land by 2050.
The averaged slope steepness for urban land would increase from
4.87 percent in 1999 to 8.32 percent in 2050 (Table 3), indicating
many woody area would be converted into urban use.
Under the second scenario, by 2050, the total urban area would be
906,134 ha., or approximately 55.40 percent ofthe entire modeled
area. The total net urban increment would be 413,003 ha., or 22.2
ha. per day, indicating an increase of 84 percent between 1999 and
2050. Apparently, the magnitude of urban growth as projected
under this scenario has been substantially suppressed. The mean
1229
